Can Smart Watches Measure Blood Pressure? What The Sensor Actually Does

Take the four categories one at a time, because lumping them together is how people end up disappointed.

Ordinary smart watches and fitness trackers that use only an optical sensor do not measure blood pressure and, with today’s sensor hardware, cannot. They measure pulse rate very well, pulse rhythm reasonably well, and blood oxygen approximately. Blood pressure is not on that list, which is why a device can tell you your heart beat 61 times a minute while having no idea whether that heart was pushing against 110 or 170. If you have ever wondered whether pulse and pressure are the same quantity, the short answer is that beats per minute and blood pressure are different measurements that can move in opposite directions.

A handful of watches do show a blood pressure number. Every one of them, without exception so far, asks you to calibrate the watch against a real arm cuff first, and then asks you to repeat that calibration every few weeks. That requirement is a confession. A device that could measure pressure would not need to be told your pressure. The calibration is the device borrowing accuracy from the cuff you already own and then slowly giving it back.

Watch-shaped devices with an actual inflatable cuff inside the strap are a real and legitimate category. They are not using optical tricks at all. They are a wrist blood pressure monitor with a screen and a watch face, and they inflate, squeeze and hiss just like a cuff, because they are one. If you want a wrist device that genuinely measures, this is the only shape that does.

Phone apps that claim to read blood pressure from a fingertip on the camera, from the phone’s microphone, or from your face on video, are the weakest category by a wide margin. A published evaluation in a major internal medicine journal tested one of the most popular such apps against a proper reference and found that roughly four out of five people who genuinely had high blood pressure were told their reading was in the normal range. That is worse than useless. A device that misses nothing is fine, a device that misses everything is at least honest, but a device that reassures the people who most need to act is actively dangerous.

You need this section. Not because the mechanics are fascinating, though they are, but because you cannot tell whether a wearable is cutting a corner unless you know which corner exists.

Blood pressure is force per unit area against the wall of an artery, and it changes continuously through every heartbeat. The only way to read it directly is to put a catheter into an artery, connect it to a pressure transducer and watch the waveform on a monitor. That is done in intensive care and in operating rooms and it is the reference against which everything else is judged. Nobody is doing that at home.

Everything outside a hospital works by a trick: stop the blood, then let it back in, and note the outside pressure at the moments when the flow changes character. That is what a cuff is for. Wrap it around the upper arm, inflate it above your systolic pressure, and the brachial artery collapses. No blood gets through. Now let the air out slowly. At the instant cuff pressure drops just below the peak pressure in the artery, the highest spike of each heartbeat punches a little gap open and a jet of blood squirts through a nearly closed tube. That is your systolic number, and you are reading it off the pressure gauge on the cuff, not guessing it.

The manual method: listening for the moment flow changes

With a stethoscope over the artery, that jet is audible. It makes a sharp tapping sound, the first Korotkoff sound, and the pressure on the gauge when tapping begins is systolic. Keep deflating. The sounds change character, get muffled, and then vanish entirely, because the artery is no longer being squeezed enough to disturb the flow. The pressure at which sound disappears is diastolic. It is a beautifully direct method and it is still the reference standard for validating home devices. If you want the full technique, our walkthrough on how to check manual blood pressure covers the deflation rate, the cuff placement and the two mistakes that ruin most attempts.

The automatic method: feeling the artery drum against the cuff

Your home monitor has no ears. What it has is a sensitive pressure transducer listening to the air inside the cuff. As the artery opens and slams shut against the cuff wall with each beat, it pushes tiny pressure ripples into that trapped air. The machine records the size of each ripple against the cuff pressure at that moment, building a curve of oscillation amplitude. The peak of that curve, where the artery is flapping most freely, happens almost exactly at your mean arterial pressure.

Here is the part that surprises people. An automatic cuff does not measure systolic and diastolic directly at all. It measures mean arterial pressure, and then derives the other two from fixed amplitude ratios in a proprietary algorithm. The one number an oscillometric monitor is most confident about is the one it never shows you, which is why running your reading through a mean arterial pressure calculation is a genuinely useful habit rather than an academic exercise.

A reading of 124/78 gives a mean arterial pressure of 78 + (124-78) / 3, which is 78 + 15.3, so about 93 mmHg. Normal sits roughly between 70 and 100, and 60 is the rough floor below which organs stop being reliably perfused. The pulse pressure, which is simply systolic minus diastolic, is 46 here, close to the typical 40. If you want the arithmetic and what the number means clinically, we walk through how to find mean blood pressure in detail elsewhere.

So even the cuff involves an algorithm. That is worth being honest about, because it is the strongest argument a wearable manufacturer has. The difference is one of degree so large it becomes a difference in kind. The cuff’s algorithm operates on a signal physically generated by the pressure it is trying to report, and it always has a fresh, live pressure reference in the form of the cuff gauge. The wearable’s algorithm operates on a signal generated by light, and its pressure reference is a memory of a cuff reading taken weeks ago. Everything in this article follows from that sentence.

Turn your watch over. Those little glowing lights are usually green, sometimes with infrared and red neighbors, and next to them sits a photodiode. That assembly is a photoplethysmograph, mercifully shortened to PPG, and it is the sensor doing nearly all the work in every wrist wearable on the market.

The physics is simple. Green light is absorbed strongly by hemoglobin. When your heart contracts, a pulse of blood swells the tiny vessels just under your skin, more hemoglobin sits in the light path, and less light bounces back to the photodiode. When the pulse passes, the vessels shrink slightly and more light returns. Sample that reflected light a hundred times a second and you get a waveform that rises and falls with every heartbeat.

What that waveform contains is genuinely rich. Beat timing, and therefore heart rate and the beat-to-beat variability that sleep and recovery scores are built from. Waveform shape, including how steeply the pulse rises, where the secondary bump from the reflected wave lands, and how the tail decays. Relative changes in blood volume in the skin. With two wavelengths, an estimate of oxygen saturation from the differing absorbance of oxygenated and deoxygenated hemoglobin.

What that waveform does not contain is pressure. There is no pressure sensor on the back of your watch. Nothing in the optical path responds to force per unit area. The device is measuring how much light came back, and light coming back is a function of blood volume in the skin, which is a function of pressure, vessel stiffness, vessel diameter at rest, skin thickness, skin tone, temperature, hydration, how tight the strap is, whether your wrist is above or below your heart, and how much your arm just moved. Pressure is one input among a dozen, and the sensor cannot tell you which input changed.

Why this matters more than it sounds

Consider what happens when you step outside on a cold morning. Your peripheral vessels constrict to conserve heat. Less blood in the skin, a smaller optical signal, a different waveform shape. Your actual blood pressure has probably gone up slightly because of the same vasoconstriction, but the optical signal has changed for a reason the model may read either way. Now consider a hot bath: vessels dilate, the optical signal swells, and your blood pressure has genuinely dropped. Sometimes the confound and the true signal point the same direction and the model looks brilliant. Sometimes they point opposite ways.

The same ambiguity explains why wearables handle some questions well and others badly. Ask a watch whether your heart rate rose during a workout and it will answer confidently, because beat timing is unambiguous in the waveform. Ask it whether your blood pressure is high after exercise and it is guessing from a signal that exercise has scrambled in four different ways at once.

None of this is a criticism of PPG. It is a superb sensor for what it senses. The problem is entirely that a category of product decided to sell it as something it is not.

Given that the sensor cannot feel pressure, how does any watch produce a systolic and diastolic reading at all? Two main approaches, and both are more clever than they are reliable.

Approach one: pulse transit time

Your heart ejects blood and a pressure wave travels down the arterial tree, arriving at your wrist a fraction of a second later. That delay is pulse transit time. The speed of that wave depends on how stiff the artery wall is, and artery walls stiffen as the pressure inside them rises. Higher pressure means a faster wave means a shorter transit time. Measure the delay, invert the relationship, report a pressure.

To measure a transit time you need two events: when the pulse left, and when it arrived. The arrival is easy, the PPG sees it. The departure is the problem. Some devices use the electrical signal of the heartbeat from an ECG electrode, which is why a few blood pressure features insist you press a finger on the crown or the case. But the electrical signal fires before the heart muscle has finished contracting, and the gap between electrical firing and mechanical ejection changes with how hard your heart is squeezing. Adrenaline shortens it. That gap is folded into the measurement and misread as a pressure change. Other devices use two optical sensors at different points and time the wave between them, which sidesteps the electrical problem but needs the two points to be far enough apart to measure a delay of a few tens of milliseconds through a moving, squishy wrist.

Approach two: pulse wave analysis

Instead of timing anything, look at the shape of a single pulse. A stiff, highly pressurized arterial system produces a pulse with a fast upstroke, an early reflected wave riding on top of the main peak, and a particular decay. A relaxed system produces a rounder shape with a late, small reflection. Extract dozens of shape features, feed them to a regression model or a neural network trained on thousands of paired PPG signals and cuff readings, and out comes a number.

This works better than it has any right to, for one uncomfortable reason. The model has learned what an average person’s waveform looks like at an average blood pressure, and most of the time most people are near average. A model that always answered 126/80 would be right within 10 mmHg for a large fraction of the population. Prediction accuracy that comes from the population distribution rather than from your artery is exactly the illusion that makes these features feel convincing at first.

Why both approaches need you personally

Whichever method a device uses, the relationship between the signal and your pressure is not universal. Two people can have identical pulse transit times at wildly different pressures because one has stiff arteries and the other has soft ones, or because one has thicker tissue between sensor and vessel. The model can learn the shape of the relationship from a population but not its position for you. So it asks you for a few cuff readings, fits your personal offset and slope, and from then on reports pressures relative to that anchor.

That anchoring is what makes cuffless blood pressure work at all, and it is also the thing that quietly falls apart. Which brings us to the section that matters most.

Calibration is the step where a cuffless device stops being a physics instrument and starts being a bookkeeper. You take three cuff readings, type them in or let a paired monitor send them over, and the watch fits your personal curve. From that moment the watch is not reporting your blood pressure. It is reporting the calibration blood pressure, adjusted by whatever it thinks has changed in your waveform since.

Four separate things then start pulling that anchor loose.

1. Your vessels change tone constantly

Arterial stiffness is not a fixed property of your body. Smooth muscle in the vessel wall contracts and relaxes minute to minute under the control of your nervous system. Stress raises tone. Sleep lowers it. Standing changes it within seconds. A large meal redirects flow to the gut. Caffeine affects blood pressure partly by altering that tone, and so does nicotine, alcohol and a cold room. The model calibrated against your vessels in one state, and then your vessels moved.

2. Your blood pressure itself has a daily rhythm

Pressure follows a circadian pattern with a morning surge, a plateau, an evening decline and a nocturnal dip of 10 to 20 percent in most people. Calibrate at nine in the evening, use the device at seven the next morning, and you are asking the model to extrapolate across the largest natural swing in your day. This is exactly why the timing rules in our guide to the best time to check blood pressure exist for cuffs too, and why they matter double for a device that is extrapolating.

3. Slow physiological change

Weight shifts, fitness improves or fades, medication doses change, kidney function drifts, arteries stiffen with age. All of these move the true relationship between your waveform and your pressure. None of them announce themselves to the device. This is the reason manufacturers typically require recalibration every four weeks or so, and it should be read as an admission that the model’s shelf life is measured in weeks.

4. Mechanical and positional change

You tightened the strap. You moved the watch a centimeter up your forearm. Your wrist is colder today. You are lying on your side. Your hand is below your heart instead of level with it, which for a wrist device alone can shift a reading by 10 mmHg or more from hydrostatic pressure. Position sensitivity is why which arm you use and how you hold it is not a trivial detail for any wrist-worn device.

The failure mode nobody advertises

Here is the specific way a calibrated cuffless device fails, and it is not the way you would expect. It does not become randomly wrong. It becomes conservatively wrong. A model anchored to your calibration values has a strong pull back toward those values, because the waveform features it reads are noisy and the safest statistical bet is always to hedge toward the mean you gave it. The result is range compression. Your true pressure swings across 45 mmHg over a month and the device reports a gentle band 15 mmHg wide, centered on the day you calibrated.

That compression is fine on the days you are near baseline, which is most days, which is why the device feels accurate. It fails on the days you are not, which are the only days that would have changed your behavior. The device is at its worst precisely at the extremes where a reading has consequences, and it is at its best on the days when the reading did not matter. No other measurement device is graded generously for this. A thermometer that read 98.6 whenever you were healthy and 99.4 when you had a fever of 103 would be pulled from shelves.

Abstract arguments about calibration do not land until you put numbers on them. These three examples use invented but realistic figures, chosen to show the shape of the error rather than to claim any particular device performs this way.

Example one: the slow drift

You calibrate on a Sunday afternoon after a relaxed weekend. Your cuff says 124/78. The watch fits itself to that and mean arterial pressure works out at 78 + (124-78) / 3, which is 93.3 mmHg. For three weeks the watch reports numbers between 119/75 and 130/82 and you feel reassured.

What actually happened over those three weeks: a stressful project started, sleep dropped to five hours, and you stopped walking at lunchtime. Your true morning pressure climbed to 148/91. Mean arterial pressure is now 91 + (148-91) / 3, which is 110 mmHg, nearly 17 points higher than at calibration. The watch, anchored to a three-week-old memory and seeing waveform changes it partly attributes to sleep debt and partly to nothing, reports 133/84. You are firmly in stage 2 territory by US criteria and the device is showing you a number that reads as mildly elevated. Whether 140/90 counts as high blood pressure is a settled question in every guideline on earth, and you never found out you were above it.

Example two: range compression

Over one month you take 60 paired readings, cuff and watch together. Your true systolic ranges from 116 to 168 depending on time of day, stress and whether you had just argued with someone. The watch reports a range of 122 to 141.

Now compute the metric that manufacturers like to quote. Average error across all 60 readings comes out at roughly 2 mmHg, because the overestimates at the low end cancel the underestimates at the high end. By that number the device looks excellent, well inside the tolerance a cuff has to meet. But look at the 8 readings where your true systolic exceeded 160. On those the device averaged 139, understating by more than 20 mmHg every time. Mean error hid a total failure to track the thing you cared about, and this is the single most common flaw in how cuffless devices get evaluated.

Example three: the acute event

You develop a severe headache and check your watch. It reports 138/86. Your true pressure at that moment is 186/118. That is above the emergency threshold of 180 systolic or 120 diastolic, and combined with a severe headache it is a reason to seek emergency care immediately, not to lie down and wait. The watch cannot rise that fast because it was never built to. Its model was trained on the narrow range of pressures people exhibit while wearing a watch calmly enough to get a clean signal, and a hypertensive emergency is outside its training distribution entirely. If you want to know what genuinely dangerous readings look like, we cover the danger level for blood pressure and the pressures associated with stroke risk separately.

This table is the quick reference. Find the thing on your wrist, your finger or your phone, and read across.

Notice the pattern down the third and fourth columns. Every device that shows a genuinely measured number needs no calibration, and every device that needs calibration is not measuring. That is not a coincidence, it is the definition.

Model names age badly, so this section deliberately talks about what the categories can do rather than what any particular watch shipped last spring. If a manufacturer announces something new next month, the questions below still apply and will still sort the real from the theatrical.

What mainstream watches genuinely do well

Continuous heart rate through the day, with resting heart rate trends that are meaningful over weeks. Irregular rhythm notification, where the watch spots an inconsistent pulse pattern suggestive of atrial fibrillation and prompts you to record a single-lead ECG. That feature is cleared as a medical device in many countries and it is one of the few consumer wearable claims with serious evidence behind it. Workout intensity zones. Sleep timing, and rough sleep staging. Blood oxygen, approximately, mostly useful for trend rather than absolute value.

Rhythm detection matters for blood pressure in an indirect way that most people miss. Automatic cuffs assume reasonably regular beats when they pick the oscillation peaks, and atrial fibrillation makes them unreliable, often needing several readings averaged or a manual check. So a watch that flags an irregular rhythm has just told you something useful about how to take your blood pressure, without measuring it. Similarly, the resting heart rate trend a watch tracks is a different quantity from pressure, though people conflate them constantly, and a low heart rate does not mean low blood pressure.

What a blood pressure feature on a watch actually is

When a mainstream watch does offer a blood pressure reading, the pattern is consistent enough to describe generically. You pair a validated arm cuff. You take three readings in one sitting and the watch learns your personal mapping. From then on you can request an estimate by sitting still for thirty seconds. Every four weeks or so the app asks you to calibrate again. The feature is available in some countries and not others, and that geography is not about hardware, it is about which regulator was willing to accept which claim.

That patchwork tells you a lot. The same watch, with the same sensor and the same software, is a blood pressure device in one country and not in another. Where the feature does ship, it is generally cleared as an aid for people who already know their blood pressure and want to track it between cuff readings, not as a tool for diagnosis. Nobody is claiming you can find out you have hypertension from a wrist optical sensor, and if a product page implies that, the fine print will contradict it.

What no optical watch can do

Tell you your blood pressure without a recent cuff reading of yours in memory. Detect a hypertensive crisis. Confirm that a reading you did not like was wrong. Replace the seven-day averaging protocol your doctor wants. Give you a number your doctor will act on. If your goal is to establish whether you have high blood pressure, or to see whether a change in treatment worked, a watch is not part of that process. A cuff and the discipline in our guide to when to take a blood pressure reading are.

Smart rings deserve their own paragraph because the marketing around them is unusually restrained and the questions people ask about them are unusually optimistic.

A ring sits on the palmar side of a finger over the digital arteries, which is arguably a better optical site than the wrist. Less motion artifact at night, less tissue between sensor and vessel, a cleaner pulse waveform. That is why rings are good at exactly the things a clean waveform gives you: night-time heart rate, beat-to-beat variability, respiratory rate estimates, skin temperature trends and sleep timing.

None of that is blood pressure. The major ring makers do not claim it is. A ring has the same fundamental limitation as a watch, no pressure sensor and no live reference, plus a smaller battery and less room for a second optical site to time a pulse wave against. If a ring ever does report blood pressure, it will be through the same calibrated inference described above and it will carry the same shelf life.

Fitness bands are simpler still. Most of them report heart rate, steps, calories and sleep, and their blood pressure story is that there isn’t one. Some companies in this space have run research studies exploring whether wrist PPG can track blood pressure, which is a legitimate thing to do and is not the same as shipping a feature. A research study is a company asking the question honestly. A product page claiming a measurement is a company answering it dishonestly.

Chest straps are the outlier worth mentioning because they are the most accurate consumer heart sensor by a wide margin and they still tell you nothing about pressure. They read the electrical signal of the heart directly rather than inferring beats from light, which makes their beat timing excellent. Electrical activity is not pressure either. A perfect ECG from a person whose blood pressure is 90/55 and a perfect ECG from a person at 190/110 can look similar, which is a compact way of understanding why how blood pressure actually works is a separate subject from how the heartbeat works.

One category people forget

Some wearables now measure things that genuinely inform blood pressure management without measuring pressure at all. Sleep duration and regularity, which matter because sleep affects blood pressure substantially. Activity minutes, because exercise lowers blood pressure in a dose-dependent way. Alcohol-related heart rate elevation, which is a decent proxy for a night that will show up in tomorrow’s reading. Used that way, a wearable is a behavior instrument rather than a diagnostic one, and it is genuinely useful.

Search any app store for blood pressure and you will find hundreds of results, many free, many with high star ratings and millions of downloads. This section is the least hedged one in the article, because the physics here is not ambiguous.

What a phone can genuinely sense from your finger

Press a fingertip over the camera lens with the flash on and the phone becomes a crude photoplethysmograph. The flash lights your finger from behind, the camera sees the light that passes through, and the average brightness of the frame dips slightly with each pulse. Sample the video at 30 or 60 frames per second and you get a pulse waveform. This is real, it works, and it is why a phone can report your heart rate with decent accuracy. Some apps also derive respiratory rate from the way the waveform envelope rises and falls with breathing, which is also legitimate.

What it cannot sense

Everything limiting a watch limits a phone more. There is no pressure sensor, no live reference, and typically no calibration step at all. A watch at least asks for your cuff readings. Most phone apps ask for your age, your height, your weight and possibly your sex, then display a blood pressure number after a few seconds of finger contact. Age, height, weight and sex predict blood pressure at a population level. What the app is showing you is a demographic average with a bit of pulse-derived jitter on top, presented in the visual language of a medical measurement.

A published evaluation in a well-known internal medicine journal put one of the most downloaded such apps head to head with proper measurement in real patients. The app’s readings clustered tightly around a normal-looking value regardless of what the person’s actual pressure was, and roughly four in five people with genuine hypertension were told they were normal. The app was eventually pulled and its marketing drew regulatory action, but dozens of similar products remain available.

The category that is worth having

There is a genuinely useful kind of blood pressure app, and it is free, and it does not measure anything. It is a logbook. You take a reading with a validated cuff, type the two numbers in, and the app timestamps it, charts it, calculates your weekly averages, separates morning from evening, and exports a clean PDF for your doctor. That is the app you want. Averages are what clinical decisions are actually made on, not single readings, and doing that arithmetic by hand is exactly the sort of chore software should absorb. Some of these apps will also compute derived values for you, though you can do the same in seconds with a mean arterial pressure tool if your app does not.

How to sort one from the other in ten seconds

Open the app store listing and read the description to the bottom. If the app measures blood pressure from a sensor, it must be a regulated medical device, and the listing will say so with a clearance or a CE mark. If instead you find a line saying the app is for informational or entertainment purposes and is not a medical device, the developer has already answered your question in writing. That disclaimer is not boilerplate, it is the legal reason the app is allowed to exist. A product that genuinely measured blood pressure would not be permitted to describe itself that way.

Two words get used interchangeably in marketing copy and they mean very different things. Getting them straight is the most transferable thing in this article, because it works on any device released in any year.

Regulatory clearance

In the United States, a device that measures blood pressure is a regulated medical device. A manufacturer cannot simply ship one. It has to demonstrate to the regulator that the device is substantially equivalent to something already legally marketed, which is the 510(k) pathway most blood pressure monitors take. Clearance means a regulator looked at the evidence and permitted a specific claim, in specific words, for a specific population. It does not mean the device is accurate for you, and it definitely does not mean every feature in the app was reviewed.

Notice the escape hatch. A product marketed for general wellness rather than for diagnosing or managing disease sits outside that regime. This is why some wearable blood pressure features are described as insights, estimates, trends or wellness indicators rather than measurements. The dispute over whether a two-digit systolic number can ever be a wellness feature rather than a medical one is live and unresolved, and at least one wearable company has been told by a regulator that showing a blood pressure value is a medical claim no matter what noun sits next to it. When you see careful language like that on a product page, read it as information rather than as legalese.

Clinical validation

Validation is separate and, for your purposes, more informative. It means an independent group ran the device against a reference standard in a defined protocol with a defined population and published the results. For cuff devices the standard protocol is well established: a specified number of subjects, a spread of arm sizes, a spread of blood pressures including genuinely high and genuinely low ones, and a simultaneous reference taken by trained observers listening with a stethoscope. Tolerances are tight, on the order of a few mmHg of mean difference with a limited spread.

Here is the important part. That protocol was designed for cuffs and it is not adequate for cuffless devices. It tests accuracy at one sitting, which a calibrated device passes trivially because you just calibrated it. It does not test whether the device still works four weeks later, whether it tracks a change in your pressure, or whether it can detect a rise as opposed to reporting your baseline back to you. Standards work specifically for wearable cuffless devices exists and is developing, and hypertension societies have published statements saying plainly that cuffless devices are not currently recommended for diagnosing or managing high blood pressure. That is the professional consensus and it is not close.

How to read a claim

Three phrases and what they actually mean. Cleared means a regulator allowed a stated claim, so go read the claim. Validated should mean an independent study against a reference, so ask which protocol and where it was published. Clinically tested means nothing whatsoever and is the phrase to be most suspicious of. You can browse the wider set of blood pressure guides for the equipment side of this, and the same distinction applies to almost every consumer health sensor on the market.

Devices change faster than articles can keep up. So instead of a verdict on this year’s models, here is the checklist. Point it at anything, including a product that does not exist yet, and it will sort the serious from the decorative.

The one test you can run yourself

If you already own a cuffless device and a validated cuff, do this over a week. Take a cuff reading and a device reading within a minute of each other, twice a day, at times when your pressure genuinely differs. Early morning before coffee. Twenty minutes after a brisk walk. Late evening. Right after something stressful. You want variety, because agreement at rest proves nothing.

Then look at two things separately. First, the average gap between the two, which tells you about offset. Second, and far more revealing, whether the device moved in the same direction and by a similar amount when the cuff moved. If the cuff swung 30 points across your week and the device swung 8, the device is not tracking you, it is reciting your calibration. That is a five-minute-a-day experiment that will tell you more than any review.

It would be unfair to leave you thinking cuffless blood pressure is a scam category. It is a hard engineering problem being worked on seriously, and some of the approaches are more promising than wrist optics. None of them are in a mainstream watch today.

Applanation tonometry

Press a pressure sensor against a superficial artery, usually the radial artery at the wrist, with just enough force to flatten one side of the vessel. When the wall is flattened, the tension in the wall no longer contributes, and the sensor reads the pressure inside directly. This is real pressure sensing, not inference, and it has been used clinically for decades to derive central aortic pressure waveforms. The problem is that it demands precise sensor placement over a specific artery and a stable hold-down force, which a consumer strap that shifts when you type does not provide.

The volume clamp method

A small finger cuff inflates and deflates hundreds of times a second under servo control, holding the blood volume in the finger constant. The pressure required to do that tracks the arterial pressure beat by beat. It produces a continuous waveform that is genuinely close to an arterial line, and it is used in operating rooms. It is also a cuff, just a very fast small one, and it makes your finger uncomfortable within an hour. Cuffless it is not.

Ultrasound and bioimpedance

Small ultrasound patches can image an artery and measure how much it distends with each beat, which combined with pulse wave velocity gives a route to pressure that is more physical than optical inference. Bioimpedance approaches send a tiny current through tissue and measure the impedance change as blood volume shifts, which is a different signal from light but a similar inference problem. Both are active research areas. Both need power, contact quality and computation that a watch does not currently have.

Better models on the same optics

The most likely near-term progress is unglamorous: better signal processing, multiple wavelengths, multiple sensor sites on one device, and machine learning models trained on much larger and more diverse datasets including people whose pressure genuinely varies. This may push cuffless estimates from useless to useful for tracking trends. It will not remove the need for calibration, because the fundamental problem is that light does not sense force, and no amount of training data changes that.

What would convince a skeptic

A device that keeps a mean error within a few mmHg for eight weeks without recalibration, in a population including untreated hypertension, while correctly tracking induced pressure changes of 30 mmHg or more, validated by researchers with no financial stake. When that paper appears, the field will have arrived. Until then, everything on your wrist is an interesting signal and the cuff is the measurement. If you want to keep an eye on the tools side of this, the health calculators collection is where the arithmetic side lives.

If you came here because you wanted a wrist device to tell you your blood pressure, this is the part that replaces it. The good news is that the alternative is cheap, takes four minutes a day for a week, and produces something a doctor will actually act on.

One more option deserves a mention because it outranks everything above. Twenty-four hour ambulatory monitoring, where you wear a cuff that inflates automatically through the day and night, is the closest thing to a diagnostic standard. It catches the white coat effect, the masked hypertension your office reading hides, and whether your pressure dips at night. If your home and office numbers disagree, that is the test to ask about.

These come up constantly, and most of them are reasonable mistakes rather than foolish ones.

No wearable should ever be the thing that decides this for you. If any device, optical or cuffed, shows something alarming, confirm it with an arm cuff and act on the confirmed number.

For context on the thresholds, the US categories set by the ACC and AHA in 2017 call normal below 120 and below 80, elevated 120 to 129 with diastolic under 80, stage 1 hypertension 130 to 139 or 80 to 89, and stage 2 at 140 or above or 90 or above. European and UK guidelines and the WHO still begin hypertension at 140/90, which means a reading in the 130s can be labeled differently depending on where you live. That difference matters for whether you get a diagnosis, and our breakdown of whether 130/80 is high covers it properly. Low readings have their own thresholds, covered in whether 90/60 is too low.

Symptoms that would take you to an ER do not become less serious because a watch says you are fine. That is the whole reason this article exists.

Medical disclaimer

External references